Energy Dispersive, X-ray determination of elements in a sample, independent of

their chemical form. It is built on the fact that elements that

Fluorescence Analysis are irradiated with high-energetic X-rays have a certain
probability of emitting characteristic X-rays, the energies of
which are unique for each element. In the energy-dispersive
Peter Wobrauschek and Christina Streli (ED) systems, the emitted X-rays are detected via their
Atominstitut, Vienna University of Technology, energies. The use of the EDXRF technique has accelerated
Vienna, Austria since the 1960s as a result of the development of liquid
nitrogen cooled solid-state detectors, nuclear electronics,
Eva Selin Lindgren and small computers. Nowadays compact light-weight
electrically cooled detectors are available, which, together
Rector’s Office, University of Borås, Borås, Sweden with air-cooled low-power X-ray tubes, are perfectly suited
for handheld spectrometers. EDXRF is multielemental
and nondestructive and can be applied to large as well as
1 Introduction 1
small samples of different composition and character. If
2 Physical Principles 2 conditions are optimized, minimum detection limits can
2.1 Interactions of X-rays with Matter 2 be below the nanogram absolute or micrograms per gram
2.2 Principles in Quantitative Evaluation of concentration level for small laboratory instruments and
Element Concentration 3 into the femtogram or nanograms per milliliter region
2.3 Interelement Effects 4 for more advanced instrumentation (total reflection X-ray
2.4 Dependence of Detection Limits and fluorescence (TXRF), synchrotron radiation).
Penetration Depth on X-ray Energy EDXRF spectrometers exist at many degrees of sophis-
and Source Characteristics 5 tication, ranging from advanced laboratory instruments
3 Energy-Dispersive X-ray Fluorescence to small portable instruments for field observations. They
Instrumentation 6 can be designed for analysis of bulk material or for scan-
ning and elemental mapping of small areas. The use of
3.1 General Features 6
X-ray optics like single or polycapillaries and curved
3.2 X-ray Sources 6 mirrors in Kirkpatrick–Baez geometry lead to X-ray spot
3.3 Energy-dispersive X-ray Fluorescence sizes of micrometers and below ideally suited for micro-
Spectrometer Designs 8 XRF. Mapping with high spatial resolution is thus possible
3.4 Detector Characteristics and New opening investigations down to cell dimension.
Developments 9 Typical applications for EDXRF are analysis of
3.5 Micro-X-ray Fluorescence 10 agricultural material, medical samples, archaeological
3.6 Portable Systems 10 and historical objects, painting and fine art objects, and
environmental samples such as soil, ores, water, and
3.7 Combination with Other Techniques 10
aerosol particles.
4 Sample Handling and Quantification in
Practice 10
4.1 Sample Preparation 10
4.2 Quantitative Analysis of Major or/and
1 INTRODUCTION
Trace Constituents 12
4.3 Surface Analysis and Microanalysis 14 EDXRF is an analytical method for qualitative as well
as quantitative determination of elements in a sample,
5 Conclusions: Needs for Future
independent of their chemical form. It is built on the
Developments 14
fact that elements that are irradiated with high-energy X-
Acknowledgments 15 rays have a certain probability of emitting characteristic
Abbreviations and Acronyms 15 X-rays. The energy, E, of this radiation is uniquely
Related Articles 15 dependent on the element, which emits the characteristic
References 15 X-rays by Moseley’s law, E = K(Z − s),(1) where K and
s are constants that vary with the spectral series and Z
is the atomic number of the element. In ED systems, the
emitted X-rays are detected by their energies, whereas
Energy-dispersive X-ray fluorescence (EDXRF) is an in wavelength-dispersive X-ray fluorescence (WDXRF)
analytical method for qualitative as well as quantitative systems, the physical equivalent to energy, the wavelength

Encyclopedia of Analytical Chemistry R.A. Meyers (Ed.)

Copyright  2010 John Wiley & Sons Ltd
2 X-RAY SPECTROMETRY

of the characteristic photon is measured. This statement
can be explained by the simple relation
E = hν, λ.ν = c > E = hc/λ > E(keV)
= 1.2396/λ(nm)
from which it is seen that the wavelength and energy of
the X-rays are reciprocal to each other.
WDXRF utilizes the properties of single crystals to
diffract X-rays of a specific wavelength under a specific
angle (Braggs law) and the intensity of the diffracted
radiation is counted by a simple radiation counter e.g.
Geiger–Muller counter. As characteristic radiation for
one element has a fixed well-defined value λ only at
specific angles, a high intensity of the refracted radiation
is measured. So using a goniometer with the Bragg crystal
as the dispersive element in its center and a detector at
the circumference, a continuous rotation will result in
intensive peaks whenever the condition of Braggs law
2d sin θ = n λ
Owing to the heavy weight of the components, there are
no portable WD spectrometers available.
The use of the EDXRF technique has accelerated
since the 1960s as a result of the development of solid-
state detectors, nuclear electronics, and small computers.
(1)
Today about 14 000 EDXRF spectrometers are in use,
many of which belong to research laboratories. This
number is rapidly growing because handheld portable
spectrometers combining new detector technologies for
charge collection and cooling are available. The compo-
sition of fine art objects, scrap metals, wall painting,
or environmental samples can be analyzed on the spot,
within seconds. A vast number of scientific publications
exist on the subject of EDXRF and its applications.
Authoritative textbooks on the principles and instrumen-
tation for EDXRF are those of Bertin,(2) Jenkins et al.,(1)
Van Grieken and Markowicz,(3) and Beckhoff et al.(4)
2 PHYSICAL PRINCIPLES
(2)

is fulfilled. Theta θ is the refraction angle, λ is the charac-
teristic wavelength, and d is the distance between lattice
planes of the single crystal. By continuous rotation of the
detector from 0 to 180° , angle 2θ, while the crystal is in
position θ, collects the photons reflected. This sequential
measurement leads to a spectrum ‘‘intensity versus wave-
length’’ allowing determination of elements and concen-
tration in the sample because the intensity is proportional
to the concentration of the respective element.
In the case of EDXRF, the detector crystal, made of
lithium-drifted silicon Si(Li) or hyperpure Ge, takes on
two tasks together, namely, dispersion according to the
energy and counting of the photons of individual energy.
This collection process results in a spectrum ‘‘intensity
versus energy’’ of the characteristic radiation emitted
from the sample. More details about energy-dispersive
detectors are presented later. Typical applications
for EDXRF are analyses of agricultural material,
medical samples, archaeological and historical objects,
and environmental samples such as soil, water, and
aerosol particles. EDXRF working is simultaneous,
multielemental, and nondestructive, and can be applied
to small samples (about 0.1 mg). For scanning objects and
mapping small areas in a specimen, EDXRF instruments
with narrow beams of X-rays can be used.
WDXRF is widely used for routine analysis and several
hundred thousand instruments are distributed in steel
production, cement industry, geologic sample analysis,
worldwide. In scientific research, the high resolution is
appreciated allowing precision measurements in cases
where the line overlaps represent a problem in EDXRF.
2.1 Interactions of X-rays with Matter
When a beam of X-rays (originating e.g. from X-ray
tubes) impinges on a sample, it will interact with
the atoms in the sample by three processes, namely,
the photoelectric effect and coherent (Rayleigh) and
incoherent (Compton) scattering. For the photoelectric
effect to occur, the energy of the impinging photons has
to be large enough to create a vacancy in one of the
shells of the studied elements. The photoelectric effect
may result in emission of characteristic X-rays, but once
a vacancy has been created in an inner shell the atom
can also de-excite by emission of Auger electrons. The
probability that characteristic X-rays will be emitted –
and not an Auger electron – varies from one element
to another and is described as the fluorescence yield.
For elements of low atomic number, emission of Auger
electrons dominates, whereas emission of characteristic
X-rays is more likely for elements of high atomic number.
A sketch of the XRF process leading to emission of a
characteristic X-ray is shown in Figure 1. On the basis of
Moseley’s law the wavelength or energy of the emitted
radiation is characteristic for the element and thus the
element can be identified.
The scattering processes are of two types, incoherent
and coherent. The probabilities of both kinds of scattering
vary with photon energy and the composition of the
sample. The scattered X-rays are an important part of
the background radiation on which the characteristic
peaks are superimposed. Figure 2 shows a typical X-ray
spectrum with characteristic and scattered radiation.
ENERGY DISPERSIVE, X-RAY FLUORESCENCE ANALYSIS 3

Characteristic Incident monochromatic or contains a range of different energies.

X-ray photon Polychromatic X-ray sources can, however, be treated
Ei = EK − EL E0 > EK as a sum (or an integral) of monochromatic sources,
and the discussion below is therefore limited to the case
of monochromatic excitation. For this case, with further
limitations to homogeneous samples and in the absence of
interelement effects, a simple relationship can be obtained
between the mass thickness (or areal density), mi , of a
specific element and the measured intensity Ii of its
characteristic X-ray peak:
Photoelectron
EPE = E0 − EK Ii = Io Ki mi Di (3)

Figure 1 Schematic drawing of the XRF process in an atom.
An incident photon of energy E0 transfers its energy to an
electron in the K-shell. The electron is expelled from the atom
and leaves a vacancy. An electron from the L-shell can transit
to the K-shell to fill the vacant site. The difference in binding
energy between the two shells, EK − EL , is used by the atom to
emit a characteristic X-ray of energy E. Note that the distances
between the atomic shells are not drawn to scale but shown
schematically only.
2.2 Principles in Quantitative Evaluation of Element
Concentration
The quantitative evaluation of element concentra-
tions in EDXRF depends on whether the source is
where I0 is the intensity of the primary beam of energy E0 ,
which, in the case that the source is a secondary-target X-
ray tube arrangement, is influenced by the anode material
and operating voltage and current; Ii is the intensity of
characteristic X-rays (K- or L-radiation) of element ‘‘i’’
of energy Ei .; Di is a correction factor for attenuation
of both the incident X-rays and the characteristic X-
rays of element ‘‘i’’ in the sample; and Ki is a factor that
depends on the setup of the instrument and basic physical
processes.
Ki thus depends on instrumental factors and on
fundamental physical parameters related to the energies
of the incident and characteristic X-rays. The important
parameters to be considered are

MoKa
inc.
coh.
MoKb
inc.
coh.
Ca

K
He Sr
Pb
As
Zn Pb Rb
Mn Br

P S Ar Cu

Cl

Cr Ni

0 100 200 300 400

Channels
Figure 2 X-ray spectrum of a biological standard reference material, SRM1571 (National Institute of Standards and Technology
(NIST)), ‘‘orchard leaves’’, measured in a secondary-target EDXRF spectrometer of the geometry of Figure 3. The sample thickness
is 30 mg cm−2 . The excitation conditions were as follows: tube voltage 55 kV and current 10 mA. The characteristic Kα lines are
denoted by chemical symbols, except for Pb, for which both Lα and Lβ are marked. Note the high intensities of the coherently and
incoherently scattered Kα and Kβ radiation of the secondary target, which, in this spectrum, is Mo. The X-ray spectrum has been
recorded with an Si(Li) detector.
4 X-RAY SPECTROMETRY
• the geometry of the XRF spectrometer;
• the attenuation of the exciting and characteristic
radiation along the beam paths and in the detector
window;
• the detector efficiency for radiation of energy Ei ;
• the probability of producing a vacancy with incident
radiation of energy E0 in the relevant shell of element
‘‘i’’;
• the fluorescence yield of element ‘‘i’’.
Note, however, that the essential feature of Equa-
tion (3) is that – under the assumptions of homogeneity
and absence of interelement effects in the sample – the
factor Ki is independent of the mass thickness, mi , of
the analyzed element and of the attenuation correc-
tion, Di .Ki can be determined through calibration of
the instrument with suitable calibration standards under
well-defined conditions.
The amount of analyte is often written as
m i = Wi m (4)
where Wi is the weight fraction mi is the mass thickness
of element ‘‘i’’ in the sample, and m is the sample mass.
The attenuation correction, Di , in Equation (3) has the
form given in Equation (5):
1 − exp −[(µ(E0 ) csc ϕ1 + µ(Ei ) csc ϕ2 ) · m]
Di = (5)
[µ(E0 ) csc ϕ1 + µ(Ei ) csc ϕ2 ] · m
In this expression, µ(E0 ) and µ(Ei ) are the mass
attenuation coefficients for the exciting and characteristic
radiation of energies E0 and Ei , respectively; they can
be expressed as weighted sums of the individual mass
attenuation coefficients for all elements in the sample.
ϕ1 is the glancing angle between the exciting radiation
and the sample surface plane and ϕ2 the glancing angle
between the characteristic radiation and the sample
surface plane. For two special cases, the attenuation
correction and thus Equation (3) simplifies tremendously,
namely, for ‘‘thin’’ and ‘‘infinitely thick’’ samples.
In some cases, it is also necessary to take into
account the fact that the characteristic radiation produced
by an element as well as the scattered radiation can
cause fluorescence of another element (interelement and
related effects, as discussed below). This effect leads to the
enhancement of one spectral line, the absorption already
having been taken into consideration. Furthermore, to
perform quantitative evaluation, one needs to know
whether impurities are present in the spectrometer
material and the properties of the detector.(1 – 3)
Equations (3) and (4) require the net peak areas
of the characteristic peaks in the X-ray spectrum to
determine element concentrations in a sample. This can
be done automatically by means of computer programs.
In evaluation of net peak areas in a spectrum, the problem
of interference between spectral lines will, however, have
to be addressed. In a number of cases, the K-lines from
one element will overlap or coincide with the L-lines
of another element, as is illustrated in Figure 2 by the
overlap of As Kα with Pb Lα. In other cases, the Kα
from one element will coincide with the Kβ from another.
By knowing relative ratios, Kα/Kβ, Lα/Lβ, it is possible
to resolve the overlapping peaks. This problem is not
so severe in WDXRF as the crystal resolution is in the
range of a few electron volts at 5.9 keV, whereas the
energy resolution of an ED detector is, for this energy
value, at its best around 130 eV at 5.9 keV. However,
this procedure of spectrum deconvolution will reduce the
precision of the measurement to an extent determined by
the statistics of the measurement.
2.3 Interelement Effects
In a sample composed of many elements, each element
will interact with the radiation from all other elements. If
the matrix elements have characteristic radiation that is
on the short wavelength side of the absorption edge
of the element of interest, element ‘‘i’’, conditions
may be favorable for the enhancement of the lines
of element ‘‘i’’. The combined effects of the matrix
elements on each other are often called interelement
or absorption–enhancement effects. Enhancement effects
are especially important for cases in which the analyte is
present at trace level and the enhancing elements at
percentage levels. In order to illustrate this, consider
the determination of one element, ‘‘i’’, by measuring its
Kα radiation in a matrix consisting of lighter as well as
heavier elements. In this case, the lighter elements will
not be able to enhance the characteristic Kα radiation
of the element ‘‘i’’ because their characteristic radiation
will not be energetic enough to create the necessary
vacancy in the K-shell of element ‘‘i’’ (see Figure 1).
Thus, it is necessary to look at the effects of the heavy
elements in the sample. If their Kα energies are lower
than that of the absorption edge for the K-shell in
element ‘‘i’’ they will not cause enhancement. If, on
the other hand, their Kα lines have higher energies than
that of the K-shell absorption edge they will enhance
the Kα radiation of element ‘‘i’’. However, the general
problem of determining whether or not enhancement
will occur is not so difficult for homogeneous samples,
since absorption edge energies for different shells and
characteristic X-ray energies for all elements are known
and readily available in tabulated form in textbooks or
from the web.
A situation similar to the interelement effect occurs
when the backscattered exciting and/or fluorescent
ENERGY DISPERSIVE, X-RAY FLUORESCENCE ANALYSIS 5

radiation excites the elements in a thick sample. These Table 1 Depth of penetration in micrometers of characteristic
effects are of relatively large importance for analysis of radiation from sulfur, chromium, and yttrium in a matrix of
trace elements embedded in organic matrices, since the iron (density 7.87 g cm−3 ) and graphite (density 2.25 g cm−3 ).
The penetration depth for reduction of X-ray intensities to
medium-weight and heavy elements are usually present in two-thirds of the original has been calculated according to
low concentrations and the incoherent scattering in these the formula I = I0 exp(−µρ x2/3 ). The elements that generate
cases is dominating.(5) (Compare the spectrum from the characteristic radiation are assumed to be present at trace levels,
organic specimen as given in Figure 2.) This fundamental which is why the density of the matrix is not affected
problem described above is the same for EDXRF and Characteristic Matrix Mass x2/3 (µm)
WDXRF and must be taken into consideration in both radiation attenuationa
techniques. (µ, cm2 g−1 )
SKa Fe 1170 0.44
2.4 Dependence of Detection Limits and Penetration CrKa Fe 119 4.3
Yka Fe 58.5 8.8
Depth on X-ray Energy and Source Characteristics Ska C 208 8.7
The intensity of a beam of X-rays passing through a Cr Ka C 15.3 120
Yka C 0.813 2200
sample is attenuated by the above-mentioned photoelec-
tric and scattering processes according to the exponential a
J.W. Robinson (ed.), Handbook of Spectroscopy, CRC Press, Boca
law: Raton, Vol. 1, 1974.

I (E) = I0 (E) exp(−[µ(E)ρx]) (6) on whether corrections for attenuation have to be made
when quantifying element concentrations. Concepts such
where I0 (E) is the original intensity of the beam; I (E) is as ‘‘thin or thick’’ samples are always seen in relation to
the intensity after the beam has traveled a distance x in the energy of the X-rays of interest.
the sample; ρ is the density of the sample; and µ(E) is the
mass attenuation coefficient of the sample material. All
these considerations are valid for both WDXRF as well 2.4.2 Minimum Detection Limit
as EDXRF.
A widely used definition for the minimum detection limits
Since the three attenuation processes that contribute to
(MDL) in EDXRF is the concentration (or, alternatively,
the total attenuation of the beam are independent of each
the amount of mass) needed for the number of counts
other, the mass attenuation coefficient can be written as
in the characteristic
√ peak of an element to be equal to
a sum of coherent, incoherent, and photoelectric mass
or larger than 3 B, where B is the number of counts √
attenuation coefficients. Each process has a functional
in the background below the peak of the element ( B
dependence of the photon energy, E, shown explicitly in
corresponds approximately to one standard deviation).
Equation (6).
Specifically

2.4.1 Penetration Depth 3 √

A relevant question in XRF analysis is how deeply the
X-rays will penetrate into the material. A quantitative
measure of this can be given in terms of the penetration
depth, which is usually defined as the path length needed
for the intensity of the X-rays to be reduced by a
given factor. Since the mass attenuation coefficients,
µ(E), in Equation (6) vary widely with X-ray energy and
material composition, penetration depths are indeed very
different for hard and soft X-rays and vary appreciably
with sample composition. Table 1 shows some typical
values for penetration depths for characteristic X-rays
of some elements embedded in an iron and carbon
matrix.(5) In this example, the penetration depth varies
by 3–4 orders of magnitude. Generally, X-rays of low
energy do not penetrate very far in a matrix of heavy
material, whereas high-energy X-rays in a light material
have high penetration. This fact has a large influence
MDL = · B (7)
k
where k is the corresponding sensitivity constant
measured as the net number of counts in the characteristic
peak per unit concentration and time.
Since the attenuation of X-rays is stronger for soft X-
rays than for hard X-rays, the background in many designs
of EDXRF spectrometers is rather flat for energies
below the scattered radiation. Best values for MDL
are usually obtained for medium-weight elements. From
Equation (7), it can be seen that the general problem of
lowering detection limits can be solved by finding ways of
increasing the intensity of the characteristic peak while
decreasing the intensity of the background radiation.
A substantial portion of the background radiation is due
to incoherent and coherent scattering of the primary
radiation in the sample itself. These scattered peaks will,
by means of secondary processes in the sample, give rise
6 X-RAY SPECTROMETRY
to a general increase in the background also at lower
energies. Furthermore, the scattered radiation will also
contribute to an increase in the background level owing to
incomplete charge collection and other processes in the
solid-state detectors. Therefore, sample properties that
influence scattering processes, for example, thickness and
average atomic number, have an essential influence on
the detection limits in an EDXRF spectrometer.
The counts in both peak and background are time
dependent, but in different ways: √ the peak increases
linearly with time, whereas B increases with the
square root of time. This means that the detection
limit will be reduced as analysis time is increased, and
the analyst can use this fact if elements present in low
concentrations are of special interest. The price paid for
improvement in MDL is, of course, a longer running time
for the spectrometer. For many laboratory instruments,
collection times in the range 200–1000 s for each X-ray
spectrum are chosen.
In practice, for bulk samples of organic material, MDLs
are of the order of 0.1 ppm for laboratory instruments for
monochromatic sources with energy matching the binding
energy for the K-shell of the element of interest.(6) In
bulk geological samples, MDLs are generally somewhat
larger. Measured as absolute amounts (for example in
mass units), the lowest MDLs can reach the picogram
region for thin samples (especially with TXRF(7) ) and the
femtogram region for microbeams that utilize a focused
synchrotron beam as source.(8)
MDLs in the 10−10 g region are reached for many
elements in thin and thick samples in laboratory EDXRF
spectrometers and MDLs with TXRF yield 10−12 g using
laboratory sources.
2.4.3 Sensitivity
A concept that is often used to describe the performance
of an EDXRF method is the sensitivity, which is defined
as the change in instrument response (net number of
counts in a characteristic peak) per unit change in the
concentration of a measured element. Thus, sensitivity in
X-ray instruments can be given as count rate per parts per
million. The sensitivity concept is useful for comparing
analyses of different elements in a given spectrometer.
It varies to a high degree with the difference in energy
between the incident and the characteristic radiation and
is highest if the energy of the incident radiation is only
slightly greater than the energy of the K-edge for the
element in question.
However, sensitivity in XRF instruments (expressed
as count rate per parts per million) is a measure of the
excitation and detection capability for each element with
a specific XRF instrument. Usually, sensitivity is not used
for comparing different instruments but, nevertheless,
is informative. For reasons of comparison, a more
appropriate parameter is the MDL achieved with each
instrument.
3 ENERGY-DISPERSIVE X-RAY
FLUORESCENCE INSTRUMENTATION
3.1 General Features
A typical feature in element analysis in practice is that
samples may contain major as well as trace concentrations
of different elements. Samples such as alloys, mineral
ores, and ceramics usually contain a number of elements,
the concentrations of which may vary by several orders
of magnitude.
In an EDXRF spectrometer, the important parts to
consider are the X-ray source, the beam and spectrometer
geometry, and the detector. An obvious goal for the
designer of XRF spectrometers is to maximize the
intensity of the characteristic X-rays from the elements
while keeping the background radiation as low as possible.
Most ED spectrometers have an elemental range from
Na to U and, if special detectors with ultrathin windows,
polymers of 300 nm supported by a Si grid are used,
the range can be extended to still lighter elements such
as Be, B, and C.(9) There are many different designs
of ED spectrometers for laboratory and field work.(10)
Combinations of WDXRF and EDXRF systems are also
used, in which the better resolution of WD instruments
is utilized for the lightest elements and the simplicity and
higher sensitivity of ED systems is utilized for elements
above Ca. The cost of an ED system is about US$ 70 000
and upward.
3.2 X-ray Sources
In photon-induced XRF, the most common sources are
X-ray tubes, which can be used directly for broadband
excitation or in combination with filters or secondary
fluorescers to obtain semimonochromatic radiation. In
some spectrometers, especially portable instruments for
field work, and for problems in which only one or a few
elements are of interest (for example, in routine process
control), radioisotopes are still useful.
3.2.1 X-ray Tubes
In X-ray tubes, the radiation spectrum is influenced
by two major processes that occur when electrons are
accelerated by the high voltage from the cathode to the
anode. One process is the creation of characteristic X-
rays in the anode material itself, which occurs when the
kinetic energy of the impinging electron is larger than
ENERGY DISPERSIVE, X-RAY FLUORESCENCE ANALYSIS
the binding energy of a particular shell in the anode
material. The other process is the creation of continuous
bremsstrahlung, which is created when the electrons are
slowed down in the anode. The bremsstrahlung spectrum
contains energies from zero up to the kinetic energy of the
electrons, which is equal to the electron charge times the
tube voltage. Accordingly, the spectrum from the X-ray
tube contains a continuous distribution overlaid with the
characteristic radiation from the anode material about
100 times higher in intensity compared to bremsstrahlung
in that energy range. By varying parameters such as anode
material and thickness, tube voltage, and current settings,
the broadband spectrum from the X-ray tube can be
optimized for the elements of interest. Some special X-
ray tubes are equipped with dual (W/Sc or W/Mo alloy)
or multiple anodes mechanically exchangeable in the high
vacuum of the tube.
Furthermore, the spot size of the electron focus on the
anode material is an important factor defining the emitted
brightness. The applicable electric power depends on the
size as the heat dissipation is the limiting factor. X-ray
tubes with standing anodes have, for Mo and W, typically
3 kW (60 kV and 50 mA) if a focus of about 0.4 × 12 mm2
is used. Broad focus tubes with 7 × 7 mm2 can dissipate
up to 40–50 kW. Rotating anodes where the anode is
formed like a wheel and rotated at 10 000 rpm allows up
to 18 kW for a 0.5 × 10 mm2 electronic focus. Low-power
X-ray tubes of 50–75 W (50 kV and 1–1.5 mA) with a
focal size of 100 µm × 100 µm emit a high flux of photons
and can be ideally combined with X-ray optical elements
to focus the beam to micrometer dimensions with a flux
of 1 × 1010 photons/s and are perfect for microanalysis
in the lab. These X-ray tubes are air cooled and of
light weight, ideally suited for the design of handcarried
portable spectrometers.(11)
As the MDL for a given element is dependent on the
intensity of the characteristic as well as the background
radiation (Equation 7), there is often a desire to use
monochromatic or quasi-monochromatic radiation. This
can be achieved by use of filters and/or secondary-target
arrangements. Filters inserted between the X-ray tube
and the sample will allow transmission of the character-
istic line of the filter but will attenuate the low-energy part
of the tube radiation by a substantial factor. Thus, filtering
may change the shape of the tube spectrum and improve
MDLs for elements lighter than the filter material. By
varying filter material and filter thickness, this method
can give a high flexibility in the use of tube excitation.
In secondary-target arrangements, the radiation from
the X-ray tube is allowed to excite a target made of a mate-
rial whose characteristic radiation (only the intensive Kα
and Kβ lines are emitted) is used as a source to excite
the sample. Figure 3 shows an arrangement of this kind,
often called triaxial geometry due to the three orthogonal
7
Secondary
target X-ray
tube
Detector
Collimator
Sample
Figure 3 Preferred geometry of a secondary-target EDXRF
spectrometer. The broadband radiation from an X-ray tube
excites a secondary target of a pure specimen, most often a
metal. The characteristic Kα and β radiation from the secondary
target is used as semimonochromatic X-ray source to excite
the different elements in the sample. A detector registers the
radiation from the sample. Note that the three X-ray beams are
mutually orthogonal and that the sample plane is horizontal.
directions of primary beam path, secondary-target beam
path, and path from sample to detector. The material in
the secondary target can be chosen so that its character-
istic radiation is just above the K-edge of the elements
of most interest in the sample. The anode material in
the X-ray tube is generally chosen to have a high output
over a wide range of energies in order to allow different
secondary-target materials to be used. Most ideal are
combinations where the characteristic radiation from the
anode is with its energy just above the absorption edge of
the target material e.g. Rh anode–Mo secondary target.
Also the tube radiation is linearly polarized after 90° scat-
tering on the secondary target in this orthogonal geometry
(all axes have an angle of 90° to each other), thus not
contributing to spectral background in the triaxial geom-
etry of Figure 3. Although the intensity of the primary
beam is reduced considerably by the secondary-target
arrangement, the technique usually gives low detection
limits for many elements. This is mainly because of the
low background in the X-ray spectra (see Figure 2). Other
techniques to modify the output from an X-ray tube
utilize low-Z elements (carbon or Plexiglas) scatterers
to polarize the entire spectrum of X-rays from the tube
or multilayers to obtain an intensive purely monochro-
matic beam, as demonstrated for TXRF spectrometers
(see Streli et al.(8) ).
For microbeam EDXRF, microfocus low-power X-ray
tubes with anode spots of the order of 100 µm2 are now
available, as are high-intensity rotating anode sources
(18 kW) with various focal spot sizes (see Beckhoff
et al.(4) ).
8 X-RAY SPECTROMETRY
3.2.2 Radioactive Sources
Radioisotopes as X-ray sources utilize the γ-rays from
the decay of the radioisotope or the characteristic X-rays
emitted from the decay product. The output intensity
is lower than that which can be achieved by an X-ray
tube, but the advantages are that the intensity of the
source is stable and no electric power supply is needed.
Radioisotopes are therefore practicable in portable field
instruments. The drawback is the limited half-lives for
some of the most commonly used isotopes (for example
2.7 years for 55 Fe with 5.9- and 6.4-keV X-rays (Mn Kα
and Kβ lines) and 1.3 years for 109 Cd with 22.2- and
25.5-keV X-rays (Ag Kα and Kβ lines) plus 88.2 keV
γ-rays). Also 241 Am with 59.5 keV γ-rays is useful for
higher-Z element excitation and this isotope has a half-
life of 433 years which is convenient. The most common
geometries of the radioisotope sources are either that of a
point source or that of an annular source. The sensitivity
and MDLs are, for some instruments, almost identical to
those for X-ray tube secondary-target arrangements.(12)
3.2.3 Synchrotron Radiation
Synchrotron radiation is the ideal source for EDXRF,
since it can be monochromatized to the desired
energy with high brilliance. Furthermore, it is linearly
polarized in the plane of the electron or positron orbit,
which is useful for decreasing scattered radiation and
thus reducing the background. Synchrotron radiation
at the National Synchrotron Light Source ((NSLS),
Brookhaven National Laboratory) with collimated and
focused X-rays of high energy has been used in
an X-ray microscope, mapping trace elements with
a spatial resolution of 10 × 10 µm2 and with MDLs
into the femtogram region for thin specimens.(13)
However, synchrotron radiation is expensive and is not
generally available in research laboratories. TXRF with
synchrotron radiation has brought a major breakthrough
in the surface analysis of Si wafers at ultratrace levels.
Detection limits of absolute 4 fg for Ni corresponding
to the more used units in surface contamination
of 5 × 108 at cm−2 were achieved in Hamburg at
Hasylab Deutsches Elektronensynchrotron (DESY) with
monoenergetic synchrotron radiation from a bending
magnet beamline.(14) Also at Stanford Synchrotron
radiation laboratory (SSRL) TXRF measurements were
performed showing the low-Z element detection capacity
with optimized excitation condition.(15)
In a critical review of synchrotron radiation total
reflection X-ray fluorescence (SR-TXRF), the advantages
and disadvantages can be found in Streli et al.(8)
Synchrotron radiation is most effective combined with
X-ray optical elements like single or polycapillaries or
Kirkpatrick–Baez optics with two mirrors leading to
sensational beam dimensions of micrometers and even
down to 30 nm. This is ideally suited for elemental analysis
of cell material and inclusions in geological formations.
Also micro-XRF in human bone, radioactive particles
and paintings were reported Janssens et al.(16)
3.3 Energy-dispersive X-ray Fluorescence
Spectrometer Designs
As already pointed out, there are three main points to
consider for achieving good performance in an EDXRF
system, namely the X-ray source, the detector, and the
design of the spectrometer. A few things of importance for
the design are mentioned here, namely, beam geometry,
choice of materials, and compactness of the design.
The problems are discussed from the perspective of
a secondary-target setup but are easy to generalize for
other arrangements.
3.3.1 Beam Geometry
A large part of the background signals in X-ray spectra
depends on primary and secondary scattered radiation in
the sample. This can be illustrated in the secondary-target
XRF arrangement shown in Figure 3. In such a system,
contributions to the background in the final spectrum are
obtained from (i) the scattered primary radiation from
the tube; (ii) the scattering of the quasi-monochromatic
beam from the secondary target; (iii) the radiation from
the sample; (iv) detector artifacts like tailing stemming
from incomplete charge collection; (v) escape and sum
peaks; and (vi) Compton backscattering from high-energy
photons leading to high background in the lower energy
region. Instrumental blank spectrum from collimators can
be a problem in the detection of ultratrace elements if the
instrument lines are the same as in the analyzed sample.
The primary radiation from the X-ray tube is very
intense and, although it will be scattered twice (in the
secondary target and the sample), it is likely to contribute
to the background unless special care is taken. The
radiation from the secondary target will be scattered
upon reaching the sample and constitutes the four large
scattered peaks shown in Figure 2. These peaks have been
observed to be responsible for part of the flat background
in the low-energy region by multiple scattering in the
sample and by incomplete charge collection effects in the
detector.
From basic physical principles, it is well known that
scattering is at a minimum when the scattering angle is
90° . Thus, it is important that the impinging and emitted
radiation form an angle of 90° with respect to each other.
This will be achieved if the secondary target and the
sample are both oriented at angles of incidence of 45° with
ENERGY DISPERSIVE, X-RAY FLUORESCENCE ANALYSIS
respect to the beam directions. Because the radiation from
the X-ray tube is scattered twice – in the secondary target
and in the sample – and since the reason for scattering
at 90° is to utilize both minimum scattering cross section
and the production of a linearly polarized beam, it is also
of importance that the geometry of the three beam paths
is a triaxial (xyz) geometry as shown in Figure 3 and not
an xyx planar geometry.(17)
3.3.2 Choice of Materials – Compact Designs
In an EDXRF spectrometer, the major part of the primary
radiation emitted by the tube and secondary target will
not only be directed toward the sample but will also
interact with the spectrometer. This interaction will take
place by means of the same processes as in the sample
itself. Different materials have different abilities to scatter
and absorb radiation, and the purity of the material will
have an influence on whether characteristic lines from
impurities in the material close to the beam paths will
pass into the detector. Thus, it is important that the
materials in objects such as filters and collimators are of
the highest possible purity.
Another important aspect to consider in an XRF setup
are the actual geometric dimensions of the various parts of
the spectrometer. Since the general goal is to have efficient
and high sensitivity in the instrument, long beam paths are
of disadvantage. In air, a long beam path for the radiation
from sample to detector will contribute to the attenuation
of characteristic radiation from the sample, and thus
significantly reduce detection efficiency for light elements
in particular. Thus, a geometry as compact as possible
with short beam paths is desirable. However, the material
in the housing of the spectrometer has to be chosen so
that unwanted radiation can be stopped efficiently. The
combination of compact geometry and high stopping
efficiency generally means that light materials such
as aluminum should be avoided or that the radiation
channels have to be clad with higher-Z materials having
good absorption properties.
Another aspect to consider in relation to Figure 3 is the
position of the sample. In many spectrometers, the sample
position is vertical, which means that samples have to be
self-supporting. By choosing an arrangement in which the
sample plane is horizontal and the detector looks down or
up at the sample, a more versatile instrument is achieved,
in which the sample can also be in nonself-supporting
forms, for example, as liquids and powders. With a
vertical sample position, such specimens are difficult to
irradiate directly by the beam. Measurements of light
elements require a special environment – either vacuum
or He flushing – to avoid absorption of the low energy
radiation. Thus vacuum chambers holding all components
of the spectrometer, in particular, sample and detector
9
snout inside, are designed for this purpose. In combination
with special detectors having ultrathin windows, elements
from Be upward can be detected. To evacuate the sample
area comprises two other advantages: the scatter of the
exciting radiation in air is avoided, which improves the
scattered background, and the Ar K-lines will not be
measured since Ar is not present in vacuum.
3.4 Detector Characteristics and New Developments
For ED spectrometers, liquid-nitrogen-cooled Si(Li)
detectors are the most widely used, having their perfor-
mance for X-rays of an energy ranging from 1 to about
25 keV and described by the efficiency of the detector.
For K-lines from heavier elements (30–100 keV) the
planar high-purity germanium detectors are preferable
owing to their better detection efficiency in this region;
the formerly available Ge(Li) is not produced any longer.
The classical LN2-cooled detector is bulky because of
the storage vessel (Dewar) for the cooling medium. An
elegant way to get rid of the Dewar is the use of elec-
trical cooling by Peltier elements. These Peltier-cooled
detectors operate not at 77 K but around 240 K, which is
reached by one- or two-stage Peltier elements working
with a power of 3–4 V and 1–2 A.
The fundamental working principle of an ED detector
is the absorption of the characteristic photons in the
intrinsic zone of the Si crystal. The energy absorbed
in the semiconductor material creates a certain number
of electron–hole pairs, linearly correlated to the energy
of the photons. These charges, created in the intrinsic
zone of the Si crystal, form a charge pulse collected at
the respective electrodes on the contact layers across
the surface of the Si crystal and are proportional
to the energy of the absorbed photon. This p–i–n
(positive–intrinsic–negative)-type structure of the Si
detector collects the charges along the parallel electric
field lines between p- and n- silicon and the contact layers
under reverse high voltage. A new idea to change the
way of collecting charges is the sideward drift.(18) In this
case, the positive electron-collecting electrode is small
and located either at the center of the crystal or slightly
eccentric at the outside of the detector structure, and
has the electric field lines parallel to the surface of the
detector crystal. According to the collection mode, these
detectors are termed drift detectors. The advantage is
high energy resolution, high throughput, and operation
at 220–250 K. The charge cloud is collected by a charge-
sensitive preamplifier and transformed into a voltage
signal that is amplified through the classical electronic
amplification chain of preamplifier and amplifier/shaper,
and converted from the analog signal into a digital
one by the ADC (analog-to-digital converter). Modern
electronics use direct conversion and digitize the preamp
10
signal immediately (DSP (digital signal processing))
leading to best possible energy resolution in the range
of 125–130 eV at 5.9 keV, which is the accepted reference
energy emitted from a 55 Fe source. Another new feature
of a modern ED detector is the extension of the detectable
range of elements to the light elements like Be, B, C, N,
O, F, Na, Mg, and Al. This is possible because, instead
of the Be entrance window, polymer structures of 300-
nm thickness supported by a Si grid (Moxtek window
trade mark (TM)) can withstand air pressure and are not
absorbing as strongly as 8-µm Be which normally is in
use. Also, the contact layers on the crystal are extremely
thin giving an adequate efficiency in the energy region of
the light elements. All measurements for light elements
have to be performed in vacuum to avoid absorption of
the fluorescence radiation on the path to the detector.
The dimensions of pin-type detectors range from 10 to
100 mm2 with a thickness of 3–4 mm, the silicon drift
detectors (SDD) have typically 10–100 mm2 area but
are between 300 and 500 µm in thickness. As a result
of the thinner crystal dimension, a poorer efficiency of
the SDD is observed. It continuously drops from about
10 keV compared to the Si pin where the decrease of
the efficiency starts roughly at 20 keV. For benchtop and
portable spectrometers, scintillation counters and gas-
filled proportional counters are sometimes used, often
in combination with filters to compensate for the lack
of energy resolution. Other types of detectors have
been developed, for example, mercury iodide detectors
at room temperature for high- and medium-Z-element
analysis.(19)
High-purity Si detectors are now able to separate
K fluorescence lines down into the 1-keV region and
compare well with compound semiconductor detectors
such as CdZnTe and HgI2.(20) An overview of the present
status of several semiconductor detectors can be found in
the article by McGregor and Hermon,(21) and Beckhoff
et al.(4)
Arrays of SiLi, GeHP, or SDD are available with
up to 100 elements, increasing the solid angle of
detection drastically – obviously extremely expensive
instruments.(22)
3.5 Micro-X-ray Fluorescence
The use of X-ray optical elements such as single or
polycapillaries, where X-ray total reflection occurs on the
inner wall side in of a thin-walled tube, on the inner wall
side X-ray total reflection occurs is leading to a focussed
beam of high intensity. A bunch of several hundred
thousands of single capillaries form the polycapillary with
intensities several orders of magnitude higher than a
pinhole of the same dimension. Scanning in steps of
micrometers according to the spot dimensions allows high
X-RAY SPECTROMETRY
spatial resolution mapping. In a particularly interesting
geometry, placing a second polycapillary in front of
the detector and aligning the system to have both foci
matching on the surface of the sample where the X-ray
beam impinges, leads to the so-called confocal geometry
with the possibility to scan layer by layer starting from
the surface and getting a 3D image at the end. Even
absorption effects can be corrected as the depth is well
known. A typical voxel size is 20 × 20 × 20 µm3 . Since
the spectra in each measured position is known, an
elemental presentation in 3D is finally possible (chemical
imaging).(23)
An excellent review of microanalysis is given in Ref. 16.
Figure 4 shows an example of 2D imaging of a bone
sample.
3.6 Portable Systems
Developments in excitation sources, detectors, and micro-
processor technology have facilitated the design of ED
configurations capable of analyzing most elements with
detection limits below the milligram per kilogram level.
In the last 20 years, portable XRF systems have also been
developed and are now commercially available, making
it possible to take the spectrometer to the sample rather
than the other way round. A detailed review has been
published by West and Potts.(24)
Special applications find the portable systems in the
field of archaeometry and cultural heritage investigations
as the technique is nondestructive. Further in-field
measurements and screening of hazardous substances
are possible with handheld systems.
3.7 Combination with Other Techniques
A comprehensive review on recent technological advan-
ces is given by Tsuji et al. in Tsuji et al.(25)
4 SAMPLE HANDLING AND
QUANTIFICATION IN PRACTICE
4.1 Sample Preparation
A typical feature in element analysis in practice is that
samples may contain high as well as trace concentrations
of different elements. Samples such as alloys, mineral
ores, and ceramics usually contain a number of elements
whose concentrations may vary by several orders of
magnitude.
EDXRF analysis can be performed on a vast number
of samples of very different composition and status.
Examples of common samples are
ENERGY DISPERSIVE, X-RAY FLUORESCENCE ANALYSIS 11

BE I
II
III
V

IV

100 µm
(a)

11400 370
Ca Sr

0 0

4000 1200
Zn Pb

0 0
(b)
Figure 4 Elemental (Ca, Sr, Zn, and Pb) 2D image of a bone sample.

• geological – for example minerals, rock, soil, and
sediments;
• biological – roots, leaves, stems, moss, lichens, algae,
fish, and animals;
• medical – blood, hair, skin, muscle, brain, teeth, bone,
enzymes, and metalloproteins;
• industrial – metals, ores, paper, and pulp;
• cultural – pottery, paintings, statues, coins, books, and
clothes;
• environmental – water, air pollutants, waste, and
pollen;
• pharmaceuticals;
• food stuff and drinking water, wine, or brandy.
There is no strict division between the different
categories, but the above list gives an idea of the
diversity of the samples concerning range of elemental
concentrations – from traces in biological material to
major components in minerals – as well as the range of
densities involved. The material may exist in different
forms, such as airborne particles, liquids, and solids.
4.1.1 Solid Samples
Homogeneous samples for which the concentrations of
the elements of interest are significantly larger than the
detection limits (usually 0.1–1 ppm) are conveniently
analyzed without any sample preparation. This is a great
advantage, since a minimum of sample treatment will
minimize the risk of contamination. Many samples belong
to this category.
Some samples may have sufficiently high element
concentrations but may be inhomogeneous and contain
grains of different sizes. In view of the poor penetration
depths for X-rays from light elements, the grain size
may affect the quantitative evaluation. For such samples,
either grinding into a fine powder or digestion and dilution
may be necessary.
4.1.2 Liquid Samples
Trace elements in water samples can be present in
different forms, for example as insoluble particulates or
as colloidal or soluble species. Suspended material can be
12
separated by filtration, and the deposits on the filter can
be directly analyzed in the X-ray spectrometer. For water,
many element concentrations are below MDLs, especially
for heavy metals. The colloidal and soluble fractions may
therefore need preconcentration, which may be achieved
by evaporation by heating or by freeze-drying. Care must
be taken, however, to avoid loss of volatile elements
in these processes. A freeze-dried residue may be acid-
digested. A number of other chemical preconcentration
procedures have been used on natural waters, snow, and
rain, with acceptable precision and recovery rates.(26)
Many of these techniques have the advantage that they
increase element concentrations, and at the same time,
giving rise to thin film samples that are easy to evaluate
quantitatively. For many fluids – and in particular if the
TXRF technique is used – the fluid can be deposited
directly onto a carrier and dried at low temperature.
TXRF shows excellent results for the analysis of Pb, Hg
in water where 10 µg L−1 MDL can be reached.
Details on sample preparation are given in Sample
Preparation For X-Ray Fluorescence Analysis
4.2 Quantitative Analysis of Major or/and Trace
Constituents
As discussed in Section 3.2, quantification involves
knowledge of some factors that are different for
each spectrometer (geometrical properties, detector and
source characteristics, etc.) and some parameters that are
dependent on physical processes. In a practical case, both
these aspects have to be taken into account.
4.2.1 Homogeneous Samples: General Approach
Samples analyzed in spectrometers with monochromatic
or nearly monochromatic X-ray sources are usually rather
easy to work with, since Equations (3–5) and (6–8)
can be used for quantitative evaluation. A convenient
approach is to calibrate the spectrometer with known
standards, which are either thin or of known mass
thickness. Calibration thus gives information on the
factors Ki for different characteristic X-ray energies,
for a certain element. These factors can be stored for
a special configuration of a given spectrometer and used
for element quantification of the unknown samples. If
an unknown sample is considered to be thin for some
elements (Di = 1), the amounts of these elements can be
readily evaluated by using Equation (3). If the sample is of
intermediate or infinite thickness, the attenuation factors
Di will have to be evaluated either from measurements
of mass thickness or from information contained in
the coherently and incoherently scattered radiation, as
discussed below.
X-RAY SPECTROMETRY
For polychromatic X-ray sources, for example broad-
band excitation from X-ray tubes, Equation (3) is not
valid. In this case, the intensity, Ii , can be described by
an integral over the distribution of excitation energies.
In practice, the approach often used is to calibrate the
instrument with standards as similar as possible to the
investigated samples.
Liquids are a special case of homogeneous samples.
For most liquid samples, quantification is facilitated by
the ability to spike the sample with internal standards
before digestion or further treatment. By choosing two
internal standards with different attenuation properties,
the attenuation for all elements in the sample can be
obtained together with information on the sample mass
thickness. In TXRF, samples are prepared on a reflector
as thin films achieved after the drying process of the liquid
solution where the addition of one internal standard was
done. The quantification is easily done with reference
to this internal standard added to the solution, using the
calibration data for each element expressed as sensitivities
achieved from a set of multielement standards with
varying concentrations. The equation reads as follows:
Cx = Ix /Sx × Sstd /Istd × Cstd (8)
where Cx denotes concentration of the unknown; Ix , the
intensity of the unknown element; Istd , the intensity of
the standard; Sx , the sensitivity of unknown element; Sstd ,
the sensitivity of standard; and Cstd , the concentration of
standard.
In this case, after the calibration of the spectrometer is
established, the sample is spiked with the known amount
of internal standard and dried on a reflector serving as
sample carrier, and after EDXRF analysis, the evaluation
of the intensities will lead immediately to quantitative
results of the unknown elements in the sample.
4.2.2 Heterogeneous Samples
For heterogeneous samples, quantitative evaluation is
more difficult, and this is particularly the case for X-rays
from light elements, for which the penetration depths
of the characteristic X-rays are small. Another problem
with heterogeneous samples is the profile of the X-ray
beam itself. In the ideal case, the beam intensity should
be constant over the beam area and drop abruptly to
zero at the edges of the beam. In reality, the beam
intensity does not behave so well, at least not for the
radiation from X-ray tubes. Owing to the geometrical
effects and different path lengths for different parts of
the radiation from the tube to the sample, the beam
may not have a homogeneous intensity distribution.
This has also been shown to be the case for secondary-
target arrangements.(27) Thus, for laboratory instruments,
ENERGY DISPERSIVE, X-RAY FLUORESCENCE ANALYSIS
knowledge of intensity variations within the beam is
essential for heterogeneous samples. Alternatively, one
can achieve good accuracy for heterogeneous samples
and beams by scanning the sample, so that effects due to
sample and beam inhomogeneity can be leveled out.
4.2.3 Methods for Quantification in Practice
Many samples can be regarded as ‘‘thin’’ for medium-
and high-Z elements, but in reality, few samples are thin
with respect to light elements of low atomic number
Z (see Table 1). Thus, one will often have to evaluate
the attenuation properties of the sample, at least for
some elements. Many methods have been developed to
assist quantification. Some of these use an experimental
approach, whereas others build on the fundamental
parameters of the physical processes involved, often
in combination with knowledge after calibration of the
instrumental performance. A few methods to illustrate
the principles in quantitative evaluations for nonliquid
samples are described.
The first method, which was one of the earliest, is
the emission–transmission method.(1,2) A precondition is
that the sample is of ‘‘intermediate thickness’’, meaning
that the intensity of the characteristic X-rays grows with
sample thickness, but it does not grow linearly. The
method is based on an independent measurement of
the factor exp[(µ(E0 ) csc ϕ1 + µ(Ei ) csc ϕ2 )m], which is
contained in the attenuation correction, Di . This factor,
which measures the sample attenuation, can be directly
evaluated from three measurements:
1. First the characteristic radiation from an element ‘‘i’’
in the sample is measured.
2. Under identical conditions, the same characteristic
radiation is measured from the combination of the
sample and a target of element ‘‘i’’ positioned behind
the sample.
3. In the third measurement, the characteristic radiation
is measured from the target of element ‘‘i’’ alone,
without the sample.
It is not difficult to show that these measurements
will give knowledge of the factor exp [(µ(E0 ) csc ϕ1 +
µ(Ei ) csc ϕ2 )m], which is needed in Equation (5). If this
factor is known, it is straightforward to calculate mi of
element ‘‘i’’ from measurement of Ii and knowledge of the
instrumental factors Ki and I0 according to Equation (3).
Another well-known method is to make use of the
information contained in the coherent and incoherent
scattering of the incident (exciting) radiation. Under
certain conditions, the intensity of the scattering peaks
or combinations of these can be used as internal
standards.(1,2,19 – 21) From Figure 2, which is an X-ray
13
spectrum from a biological sample, in which Mo has been
used as secondary target, it is seen that the incoherently
scattered radiation is very pronounced. In a biological or
other kind of light-element sample in which medium
and heavy elements are present at trace levels, the
incoherently scattered radiation is essentially a measure
of the mass seen by the impinging beam. The argument
for this is that for the light elements (except hydrogen),
e.g. carbon, nitrogen, and oxygen, which are the main
constituents in most biological samples, the mass of an
atom is closely proportional to its atomic number (Z).
Furthermore, the binding energies of the electrons in
the light elements are small compared with the energy
of the exciting radiation. Thus, the exciting radiation
essentially sees a cloud of free electrons, and the
probability of incoherent (Compton) scattering will be
proportional to the number of electrons, which, in turn,
is approximately proportional to the sample mass. This
approach is also valid for some elements of importance
in geochemical samples, such as Mg, Si, and Ca (ratios
of relative atomic mass to atomic number: 2.03, 2.01,
2.00) but less so for Na, Al, and K (ratios of whose
relative atomic mass to atomic number are in the range
2.09–2.06). For samples composed of heavy elements,
the ratios of atomic mass to atomic number will change
owing to increased neutron excess in the atomic nuclei.
Furthermore, the binding energy of the inner electrons
increases and the free electron incoherent scattering
cross section is no longer such a good approximation.
Secondary scattering effects in the material may also
appear. Several practical methods, however, have been
developed, which make use of the information contained
in both coherently and incoherently scattered radiation
for thick samples of geochemical origin. Giauque et al.(28)
and Giauque(29) have used the ratios of the incoherently
scattered radiation from Ag Kα and Kβ radiation together
with the coherently scattered Ag Kα to ascertain sample
mass thickness and photoelectric cross sections for thick
geochemical samples. Also, in the case of WDXRF,
the scattered Compton lines from the X-ray tube have
proved to be useful as internal standards in light-element
samples.(30)
The general aim of quantification in EDXRF is to
arrive at the sample composition from the information
contained in the spectrum. In an XRF spectrum, the
medium-heavy and heavy elements in the sample will
reveal their presence by their characteristic radiation
as well as by their contribution to the coherently and
incoherently scattered radiation. Thus, a first estimate
of the relative abundance of these elements can be made
directly from the spectrum. As discussed earlier, however,
this first estimate will be significantly modified by
considering attenuation and interelement effects, etc. To
evaluate the mass thickness of the light elements – which
14
do not show up by their characteristic X-rays – their
contribution to the scattering peaks can be determined
from the spectrum by subtraction of the scattering due
to the medium-heavy and heavy elements. From this
evaluation, the total mass thickness from light as well as
medium and heavy elements can be calculated.
In an iterative process, in which all the information
from the elements through their characteristic as well
as scattered X-rays is used, the concentrations can be
recalculated until the concentration values obtained agree
with the experimental spectrum.
In the practical situation, the analyst is dependent on
computer programs in which an X-ray library is stored
together with information on the actual spectrometer
(obtained by calibration samples). Spectrum evaluation
programs are available in commercial instruments
and from research groups. A more-or-less continuous
development of these programs takes place, in which
peak shapes and background profiles are described by
model functions.(31)
4.3 Surface Analysis and Microanalysis
Quantification of thin surface layers or microstructures
essentially makes use of the same fundamental methods
as discussed above, namely, adding of an internal
standard for liquids, use of external reference material,
and use of scattered radiation. For thin self-supporting
homogeneous solid surface layers, quantification can
be achieved by use of scattered radiation, calibration
constants, or fundamental parameters.
In microelectronics, layers and implants need to be
prepared in the nanometer region. In these cases, the Si
wafer, as the basic material, is the best substrate for TXRF
in order to use the angle-dependent fluorescence signal
of the layer or implant to derive information on layer
thickness, element and density – in case of implants the
concentration profile can be derived by angle scanning
from 0° to 2–3 times the critical angle. This technique
is sensitive in the nanometer range and allows excellent
insight to the new design of ultrashallow junctions. Both,
laboratory sources and synchrotron radiation can be
used for these applications, called grazing incident X-ray
fluorescence (GIXRF).(32)
In the case of heterogeneous microstructures smaller
than the X-ray beam diameter, quantification by means
of external references is difficult in spectrometers based
on X-ray tubes owing to the inhomogeneous beam profile
as discussed above, and the difficulty in positioning
the samples within the beam. In these cases, internal
standards such as the scattering peaks from the sample will
give reasonably good average values, although structures
with resolution below 10 µm are presently difficult to
study with laboratory equipment.(33)
X-RAY SPECTROMETRY
5 CONCLUSIONS: NEEDS FOR FUTURE
DEVELOPMENTS
Although EDXRF has many advantages, in that it is
multielemental, nondestructive, and needs a minimum
amount of sample treatment, it has two distinct disadvan-
tages:
• For bulk samples it is not very sensitive. MDLs for
laboratory equipment are in the order of 0.1 ppm.
Measured in mass units, it has MDLs of the order
of a fraction of a nanogram for the best cases in
realistic samples. For synchrotron radiation sources
and special designs (TXRF; grazing emission X-ray
fluorescence (GEXRF)), MDLs may be reduced by
several orders of magnitude, but not for all kinds of
samples. A substantial part of the lack of sensitivity for
bulk samples is due to effects caused by the presence
of the large scattered peaks entering the solid-state
detectors.
• In comparison with microprobes with charged parti-
cles – electrons or protons – microprobes of X-rays
with high lateral resolution are difficult to obtain. This
is because charged particles can be steered by electric
and magnetic fields, whereas X-rays will have to be
focused with mirrors or other kinds of optical devices.
The advantage of X-rays over charged particles is that
they deposit less energy into the sample and thus will
cause less heating and loss of volatile material.
Within a few years, EDXRF has mutated into a
very versatile analytical tool owing to the development
of sources, optics, and detectors. The applicability
of EDXRF has been drastically enhanced in various
fields. Quantification is no longer a big problem
since the knowledge of primary X-ray spectra and
fundamental parameters has drastically improved. The
improvement in computing capability and speed has led
to faster quantification and data evaluation. Software
packages are available as freeware, performing spectrum
deconvolution to obtain net counts for each element of the
sample and also solving the rather complicated equations
related to all interaction processes rapidly in seconds.
The elemental mapping capacity (2D and 3D) with
high power sources opens new applications in medicine,
biology, environment, and cultural heritage investiga-
tions, where quantification is not the most relevant
information.
EDXRF has reached a degree of maturity with regard
to accuracy and precision of the measurements. New
sources – X-ray optics and various excitation geometries,
like TXRF, GIXRF, and GEXRF – have shown a wide
field of applications in layer and implant characterization.
In combination with high-resolution single crystals
ENERGY DISPERSIVE, X-RAY FLUORESCENCE ANALYSIS 15

in double-crystal arrangements, energy increments of p–i–n Positive–Intrinsic–Negative

100 meV are possible; this can be used to study the SDD Silicon Drift Detectors
chemical state by measuring the fluorescence signal SR-TXRF Synchrotron Radiation Total Reflection
in dependence of the incident photon energy around X-ray Fluorescence
the absorption edge of the element considered. X-ray SSRL Stanford Synchrotron Radiation
absorption near edge spectroscopy (XANES) is used to Laboratory
determine the chemical state of the element, extended TM Trade Mark
X-ray absorption fine structure (EXAFS) is used to TXRF Total Reflection X-ray Fluorescence
determine the coordination number and the distance to WDXRF Wavelength-Dispersive X-ray
the neighboring atoms. Fluorescence
Further developments of X-ray sources like free XANES X-ray Absorption Near Edge
electron lasers X-ray free electron laser (XFEL) and Spectroscopy
Linac coherent light source (LCLS) will be soon XFEL X-ray Free Electron Laser
available for the scientific community and will require
the development of ultrafast detector technology and
electronics to make use of the high brilliant sources for
EDXRS. RELATED ARTICLES
Future development, which would imply break-
throughs on the above items, would be Environment: Water and Waste (Volume 4)
Sample Preparation for Elemental Analysis of Biological
• improvements in detector technology to minimize Samples in the Environment • X-ray Fluorescence
effects that cause increased background, for instance Spectroscopic Analysis of Liquid Environmental Samples
trapping and polarization in the detector material;
• development of laboratory X-ray lasers in order to Forensic Science (Volume 5)
obtain parallel, polarized, and intense beams. X-ray Fluorescence in Forensic Science
For both these developments it is easy to observe that
not only EDXRF but also a vast field of other analytical Peptides and Proteins (Volume 7)
methods would benefit from them. X-ray Crystallography of Biological Macromolecules

Steel and Related Materials (Volume 10)

ACKNOWLEDGMENTS
One of the authors, E. Selin Lindgren, is grateful to
Ing. Paulis Standzenieks for assistance in preparing the
manuscript.
X-ray Fluorescence Spectrometry in the Iron and Steel
Industry
Nuclear Methods (Volume 14)
Particle-induced X-ray Emission (PIXE)

Radiochemical Methods (Volume 14)

ABBREVIATIONS AND ACRONYMS Nuclear Detection Methods and Instrumentation

X-ray Spectrometry (Volume 15)

ADC Analog-to-Digital Converter
Portable and Handheld Systems for Energy-dispersive
DESY Deutsches Elektronensynchrotron
X-ray Fluorescence Analysis • X-Ray Fluorescence Anal-
DSP Digital Signal Processing
ysis, Sample Preparation For • Total Reflection X-ray
EDXRF Energy-Dispersive X-ray Fluorescence
Fluorescence • Wavelength-dispersive X-ray Fluores-
ED Energy-Dispersive
cence Analysis • X-ray Techniques: Overview
EXAFS Extended X-ray Absorption Fine
Structure
GEXRF Grazing Emission X-ray Fluorescence
GIXRF Grazing Incident X-ray Fluorescence REFERENCES
LCLS Linac Coherent Light Source
MDL Minimum Detection Limits 1. R. Jenkins, R.W. Gould, D. Gedcke, Quantitative X-ray
NIST National Institute of Standards and Spectrometry, Marcel Dekker, New York, 1981.
Technology 2. E.P. Bertin, Principles and Practise of X-ray Spectrometric
NSLS National Synchrotron Light Source Analysis, Plenum Press, New York, 1975.
16
3. R.E. Van Grieken, A.A. Markowics (eds), Handbook of
X-ray Spectrometry, 2nd edition, Marcel Dekker, New
York, 2002.
4. B. Beckhoff, B. Kanngießer, N. Langhoff, R. Wedell,
H. Wolff (eds), Handbook of practical X-ray Fluorescence
analysis, Springer, Berlin, Heidelberg, New York, 2006.
5. A.G. Karydas, T. Paradellis, ‘Calculation of the Enhan-
cement of Characteristic Radiation Due to Scattering
in Radioisotope XRF Using Exact Differential Cross-
sections’, X-ray Spectrom., 22(4), 208–215 (1993).
6. R.P. Pettersson, E. Selin Lindgren, ‘Energy-dispersive X-
ray Fluorescence Analysis’, in Surface Characterization,
A User’s Sourcebook, eds. D. Brune, R. Hellborg, H.J.
Whitlow, O. Hunderi, Wiley-VCH, Weinheim, 136–153,
1997.
7. P. Wobrauschek, Total reflection x-ray fluorescence
analysis – a review, X-ray Spectrom., 36, 289–300 (2007).
8. C. Streli, P. Wobrauschek, F. Meirer, G. Pepponi, Total
reflection X-ray fluorescence and grazing incidence X-ray
spectrometry – Tools for micro- and surface analysis. A
review, J. Anal. Atom. Spectr., 23, 792–798 (2008).
9. www.moxtek.com, 2009.
10. www.panalytical.com, www.rigaku.com, www.spectro.
com, www.edax.com, www.oxford-instruments.com,
www.thermofisher.com, 2009.
11. homepage of oxford instruments: www.oxfordxtg.com,
2009.
12. V. Valkovic, A. Markowics, N. Haselberger, ‘Review of
Recent Applications of Radioisotope Excited X-ray
Fluorescence’, X-ray Spectrom., 22(4), 199–207 (1993).
13. K.W. Jones, W.M. Kwiatek, B.M. Gordon, A.L. Han-son,
J.G. Pounds, M.L. Rivers, S.R. Sutton, A.C. Thompson,
J.H. Underwood, R.D. Giauque, Y. Wu, ‘X-ray Micros-
copy Using Collimated and Focussed Synchrotron
Radiation’, Adv. X-ray Anal., 31, 59–68 (1987).
14. C. Streli, G. Pepponi, P. Wobrauschek, C. Jokubonis,
G. Falkenberg, G. Zaray, J. Broekaert, U. Fittschen,
B. Peschel, Recent results of synchrotron radiation
induced total reflection X-ray fluorescence analysis
at HASYLAB Beamline L, Spectrochim. Acta, 61B,
1129–1134 (2006).
15. C. Streli, G. Pepponi, P. Wobrauschek, N. Zöger,
P. Pianetta, K. Baur, S. Pahlke, L. Fabry, C. Mantler,
B. Kanngiesser, W. Malzer, Analysis of low Z elements
on Si wafer surfaces with synchrotron radiation induced
total reflection X-ray fluorescence at SSRL, Beamline
3–3: comparison of droplets with spin coated wafers,
Spectrochim. Acta, 58B, 2105–2112 (2003).
16. K. Janssens, F. Adams, A. Rindby (eds), Microscopic X-
ray Fluorescence Analysis, Wiley, 2000.
17. P. Standzenieks, E. Selin, ‘Background Reduction of X-
ray Fluorescence Spectra in a Secondary Target Energy
X-RAY SPECTROMETRY
Dispersive Spectrometer’, Nucl. Instrum. Methods, 165,
63–65 (1979).
18. P. Lechner, C. Fiorini, R. Hartmann, J. Kemmer, N. Kra-
use, P. Leutenegger, A. Longoni, H. Soltau, D. Stötter,
R. Stötter, L. Strüder, U. Weber, Silicon drift detectors for
high count rate X-ray spectroscopy at room temperature,
Nucl. Instrum. Methods A, 458, 281–287, (2001).
19. J. Nissenbaum, A. Holzer, M. Roth, M. Schieber, ‘X-ray
Fluorescence Analysis of High Z Materials with Mercuric
Iodide Room Temperature Detectors’, Adv. X-ray Anal.,
24, 303–309 (1980).
20. P. Jalas, H. Sipila, S. Eranen, K. Leinonen, ‘Near Room
Temperature Energy-dispersive Detectors Processed on
High-purity Silicon’, X-ray Spectrom., 24(2), 63–69 (1995).
21. D.S. McGregor, H. Hermon, ‘Room-temperature Com-
pound Semiconductor Radiation Detectors’, Nucl.
Instrum. Methods, 295, 101–124 (1997).
22. www.canberra.com, www.ketek.net, 2009.
23. B. Kanngießer, W. Malzer, I. Reiche, ‘A New 3D Micro
X-ray Fluorescence Analysis Set-up- First Archaeometric
Applications’, Nucl. Instrum. Methods, B 211/2, 259–264
(2003).
24. Ph. Potts, M. West (eds), Portable X-ray Fluorescence
Spectrometry, RCS Publishing, 2008.
25. K. Tsuji, J. Injuk, R. Van Grieken, X-ray Spectrometry
Recent Technological Advances, Wiley, 2004.
26. B. Holynska, ‘Sampling and Sample Preparation in
EDXRS’, X-ray Spectrom., 22(4), 192–198 (1993).
27. J. Boman, P. Standzenieks, E. Selin, ‘Beam Profile Studies
in an X-ray Fluorescence Spectrometer’, X-ray Spectrom.,
20(6), 337–341 (1991).
28. R.D. Giauque, ‘A Novel Method to Ascertain Sample
Mass Thickness and Matrix Effects for X-ray Fluorescence
Element Determinations’, X-ray Spectrom., 23(4),
160–168 (1994).
29. R.D. Giauque, F. Asaro, F.H. Stross, T.R. Hester, ‘High
Precision Non-destructive X-ray Fluorescence Method
Applicable to Establishing the Provenance of Obsidian
Artifacts’, X-ray Spectrom., 22(1), 44–53 (1993).
30. H.A. van Sprang, M.H.J. Bekkers, ‘Determination of Light
Elements Using X-ray Spectrometry. Part 1 – Analytical
Implications of Using Scattered Tube Lines’, X-ray
Spectrom., 27(1), 31–36 (1997).
31. B. Vekemans, K. Janssens, L. Vincze, F. Adams, P. Van
Espen, ‘Analysis of X-ray Spectra by Iterative Least
Squares (AXIL): New Developments’, X-ray Spectrom.,
23(6), 278–285 (1994).
32. G. Pepponi, C. Streli, P. Wobrauschek, N. Zoeger,
K. Luening, P. Pianetta, D. Giubertoni, M. Barozzi,
ENERGY DISPERSIVE, X-RAY FLUORESCENCE ANALYSIS
M. Bersani, Nondestructive dose determination and
depth profiling of arsenic ultrashallow junctions with
total reflection X-ray fluorescence analysis compared to
dynamic secondary ion mass spectrometry, Spectrochim.
Acta B, 59, 1243–1249 (2004).
17
33. R.W. Ryon, H.E. Martz, J.M. Hernandez, J.J. Haskins,
E.J. Suoninen, ‘Selection of the Method’, in Surface
Characterization, A User’s Sourcebook, eds. D. Brune,
R. Hellborg, H.J. Whitlow, O. Hunderi, Wiley-VCH,
Weinheim, 14–31, 1997.